Runner job Metrics tab now shows per-job machine metrics

The Metrics tab in the runner-job dashboard now actually shows data. Previously the tab was blank because the server tried to collect machine metrics through a path that did not work on Tuist’s fleet: macOS Tart runner VMs do not expose it, and the Linux kata nodes rejected it. Sampling now happens inside the runner VM itself for both macOS and Linux, so each job reports its own whole-VM CPU, memory, network throughput, and disk usage for the full duration of the job. The change ships together with matching ingest auth, so once the updated runner images roll out every eligible runner job populates its Metrics tab; older jobs on prior images stay empty until they recycle.